THE BAMIYAN VALLEY: LANDSCAPE MODELING FOR CULTURAL HERITAGE
VISUALIZATION AND DOCUMENTATION
Armin Gruen, Fabio Remondino, Li Zhang
Institute for Geodesy and Photogrammetry, ETH Zurich, Switzerland
E-mail: <agruen><fabio><zhangl>@geod.baug.ethz.ch
Commission V - WG 6
KEY WORDS: Cultural Heritage, SPOT, IKONOS, Reconstruction, Modeling, Texture, Visualization
ABSTRACT
In this paper we present the modeling and visualization of the cultural heritage area of Bamiyan, Afghanistan. The region is situated
ca 200 km north-west of Kabul and it was one of the major Buddhist centres until the ninth century AD. The two standing Buddhas
belonged to some of the most famous Buddhist monuments world-wide. In 2001 they were destroyed through an act of vandalism by
the Taleban militia. In our previous reports we have already presented the 3D computer reconstruction of the Great Buddha of
Bamiyan while in this contribution we will demonstrate 3D visualization products generated with different software packages and a
first attempt at generating a spatial information system.
1. INTRODUCTION
The region of Bamiyan, ca 200 km North-West of Kabul,
Afghanistan (Figure 1), was one of the major Buddhist
centres from the second century AD up to the time when
Islam entered the area in the ninth century. For centuries,
Bamiyan lay in the heart of the famous Silk Road, offering
rest to caravans carrying goods across the area between
China and Western Empires. Strategically situated in a
central location for travellers from North to South and East to
West, Bamiyan was a common meeting place for many
ancient cultures.
The larger statue of Bamiyan was 53 metres high while the
smaller one measured 38 m. They were cut from the
sandstone cliffs and they were covered with a mud and straw
mixture to model fine details such as the expression of the
face, the hands and the folds of the robe (Figure 3).
Figure 3: The two big standing Buddhas of Bamiyan.
Figure 1: The map of Afghanistan and the welcome sign before
entering Bamiyan village.
In the region, many Buddha statues and hundreds of caves
were carved out of the sedimentary rock. In particular, near
the village of Bamiyan, at approximately 2500 meters
altitude, there were three big statues of Buddha carved out of
a vertical cliff (Figure 2).
Figure 2: Panorama of the Bamiyan cliff with the three
Buddha statues prior to demolition.
The invasion of the Soviet army in December 1979 started a
23-year long period of wars and barbary that left Afghanistan
as a heavily hurt and damaged country, with little hope for
quick infrastructure reconstruction, economic improvements,
poltical stability and social peace. Moreover, at the end of the
1990’s the extremist Taleban regime started an internal war
against all non-Islamic symbols. This led in March 2001 to
the complete destruction of the two big standing Buddha
statues of Bamiyan (Figure 4), as well as other small statues
in Foladi and Kakrak. In 2003, the World Heritag Committee
has decided to include the cultural landscape and
archaeological remains of the Bamiyan valley in the
UNESCO World Heritage List [http://whc.unesco.org/]. The
area contains numerous Buddhist monastic ensembles and
sanctuaries, as well as fortified edifices from the Islamic
period. The site symbolizes the hope of the international
community that extreme acts of intolerance, such as the
deliberate destruction of the Buddhas, are never repeated
again. The whole area is nowadays in a fragile state of
conservation as it has suffered from abandonment, military
actions and explosions. The major dangers are the risk of
imminent collapse of the Buddha niches (Figure 5) with the
remaining fragments of the statues, further deterioration of
still existing mural paintings in the caves, looting and illicit
excavation.
Figure 4: The explosions of March 2001 that destroyed the
Buddha statues (Image Source: CNN).
In our previous reports we have already presented the 3D
modeling of the lost Great Buddha statue and the two empty
niches by using old and recent photographs (Gruen et al.,
2004a,b) (Figure 6).
Figure 5: The two empty niches, where the Buddha once stood,
as seen in August 2003.
In this work we mainly present the modeling of the entire
UNESCO cultural heritage site using high-resolution satellite
imagery. The final goal is the production of virtual flights
over the area, the extraction of cartographic vector features
around the village of Bamiyan and the documentation of the
protected area with a tourist and cultural information system.
Figure 6: The 3D computer model of the Great Buddha, as
reconstructed from existing images (left, Gruen et al., 2004b)
and the 3D model of its actual empty niche (right, Gruen et
al., 2004a).
2. TERRAIN MODELING FROM SATELLITE
IMAGERY
For the 3D modeling and visualization of the area of interest,
an accurate DTM is required. We had the contours of
Russian maps 1:50 000 digitised, but when we tried to map
an IKONOS image onto the derived DTM we realised that it
did not fit. Aerial images were not available to us and the
idea to acquire them was unrealistic, due to the absence of
any surveying company operating in that area. So spacebased image acquisition and processing resulted as the only
alternative to the aerial photos or any other surveying
method. Nowadays space images are competing successfully
with traditional aerial photos, for the purpose of DTM
generation or terrain study in such problematic countries, as
Afghanistan is. Also the resolution and availability of worldwide scenes taken from satellite platforms are constantly
increasing. Those scenes are available in different
radiometric modes (panchromatic, multispectral) and also in
stereo mode.
2.1 The satellite images
The HRG sensor carried on SPOT-5 since May 2002 was
suitable for our tasks [http://www.spotimage.fr]. The sensor
can acquire stereo images in across-track direction at 2.5m
ground resolution in panchromatic mode. The time difference
between two successive acquisitions in stereo mode of the
same area depends on the incidence angle, with a minimum
of one day difference. The satellite flies at a mean height of
832 km, along a quasi-polar and sun-synchronous orbit.
Due to the scientific, social and cultural interest of the
project, a B/W stereo pair over Bamiyan was provided at
special conditions by the ISIS Program [http://medias.obsmip.fr/isis/]. The two scenes have been acquired on 18th
December and 19th December 2003, at 10:40 a.m. and 10:20
a.m. local time respectively. The scenes are 24000 x 24000
pixels large and cover a mountainous area of approximately
60x60 km, centered at (34º50‘ N, 68º 7` E) and (34º50‘ N,
67º 48` E). The cloud cover was zero and the ground
resolution is 2.5 m.
Furthermore, a colour Geo level IKONOS image mosaic over
the Bamiyan area was provided by Space Imaging
[http://www.spaceimaging.com] (Figure 7). The scenes were
acquired on 15th December 2001 and covers an area of 13x20
km, centered at (34º46’ N, 67º49’ E). The image size is
13957 x 21118 pixels and the ground resolution is 1m.
Figure 7: The two empty Buddha niches as observed by
IKONOS (courtesy of Space Imaging, Inc.). Left: Great
Buddha. Right: Small Buddha.
2.2 GPS measurements
For the georeferencing of the satellite images, seven GPS
points were measured (Figure 8) during our field campaign in
August 2003. Two Trimble GEO Explorer receivers logging
carrier phase data (C/A code) were used; one receiver was set
as ‘master’ (fix position) and the other served as ‘rover’. The
points were identified on the IKONOS image and then
measured in C/A mode.
After the data collection, the observations were postprocessed to generate more accurate positions. It was not
possible to perform absolute differential corrections as the
closest master station was ca 1500 km away. Table 1
summarises the GPS measurements.
scenes' orientation was performed with the help of the GCPs
measured with GPS. Table 2 shows the image orientation
results.
Table 1: GPS measurements and related parameters.
The oriented SPOT stereo pair was then subjected to the
automated DTM/DSM generation, using the module of SATPP. A 20 m raster DTM for the whole area and 5 m raster
DTM for the area covered by the IKONOS image were
interpolated from the original matching results (Figure 9),
using also some manually measured breaklines near the
Buddha cliff. The matching algorithm combines the matching
results of feature points, grid points and edges. It is a
modified version of MPGC (Multi Photo Geometrically
Constrained) matching algorithm [Gruen, 1985; Zhang and
Gruen, 2004] and can achieve sub-pixel accuracy for all the
matched features.
No. of
Points
Average
satellites
Average
PDOP
Horiz.
Accuracy
Vert.
Accuracy
7
6
2.78
0.11 m
0.3 m
Table 2: Image orientation results.
Source
IKONOS
SPOT Image Pair
RMSE
East (m)
0.56
1.22
RMSE
North (m)
0.48
2.01
RMSE
Height (m)
1.50
Figure 8: The distribution of the GPS ground control points
on the IKONOS image mosaic. In the centre the master
station position.
2.3 Scenes orientation and DTM generation
The DTM was generated from level SPOT5 1A scenes
(radiometrically but not geometrically corrected) using SAT–
PP (SATellite Precision Positioning) software developed at
our Institute. SAT-PP includes different modules for the
precise processing of high-resolution satellite image data and
allows image pre-processing, orientation, matching,
DTM/DSM generation and object extraction. For more
details, see [Zhang and Gruen, 2004; Poli et al., 2004]. The
IKONOS mosaic orientation was based on a 2D affine
transformation. On the other hand, the SPOT scenes
orientation was based on a rational function model. Using the
camera model, the calibration data and the ephemeris
contained in the metadata file, the software estimates the
RPC (Rational Polynomial Coefficients) for each image and
applies a block adjustment in order to remove systematic
errors in the sensor external and internal orientation. The
Figure 9: The SPOT derived DTM of the Bamiyan area
displayed in colour coding mode (above). The two overlaid
DTMs (below).
Figure 10: The DTM of the Bamiyan area generated from satellite images and textured with a SPOT5 (left) and a colour IKONOS
ortho-image (right).
Empty niche of the Great Buddha
Empty niche of the Small Buddha
Rock cliff with Buddha niches
The new ‘bazaar’
Shahr-i-Ghulghulah, the old Bamiyan city
Figure 11: Zoom into the 3D Bamiyan terrain model textured with colour IKONOS ortho-image. The rock cliff with the two
empty niches (above). The pyramid-type hill (Shar-i Ghulghulah) where the old city was located, the Bamiyan cliff and the new
‘bazaar’ of the village (below).
3. DTM DATA VISUALIZATION
The visualization of the produced DTM is a very important
element, as for the external world it is usually the unique and
possible contact with the 3D model. For the photo-realistic
visualization of the whole Bamiyan area, a 2.5 m resolution
B/W ortho-image from SPOT images and a 1 m resolution RGB
ortho-image from the IKONOS image were generated. The
textured 3D models (rendered with Erdas-Virtual GIS) are
shown in Figure 10 and 11, where two closer views on the 3D
IKONOS textured model of the Bamiyan cliff and the old
Bamiyan city (the pyramid-type hill to the left) are presented.
Table 3: DTM geometry and texture data
Area
[km]
49x38
SPOT
13x20
IKONOS
* uncompressed format
Grid
[m]
20
5
Points
[Mil.]
4.6
12
Geometry
file *
215 MB
550 MB
Texture
file *
432 MB
858 MB
Visualization packages and tools are available in different
forms. For DTM visualization (often called ‘geovisualization’),
more than 500 packages are available, while many other
developed software are not known and public. The packages
can be classified according to their real-time performances, the
rendering quality of geometry and texture, the anti-aliasing
function, the input data, the level of detail (LOD) properties,
etc. The Bamiyan dataset has a size of ca 640 MB (SPOT data)
and 1.4 GB (IKONOS data) and it requires very powerful
visualization software for an interactive use. We employed
Skyline (IDC AG), Erdas-Virtual GIS and ESRI ArcGISArcScene for data rendering, real-time demos and video
generation and in the following section a short comparison is
reported. All packages allows interactive or user-defined
flythrough animations.
3.1 Erdas–Virtual GIS 8.6
effect. The ‘extrude’ function can display features from 2D
vector data sources and generate lines, walls or solids in 3D.
Table 3: Comparison of geometry and texture data reduction
after the conversion to the different software internal format.
Note how Skyline can reduce the whole file size of a factor 4
SPOT
IKONOS
geometry/texture geometry/texture
Erdas
28 / 434 MB
47 / 865 MB
Skyline
28 / 397 MB
47 / 598 MB
ArcScene
27 / 432 MB
46 / 858 MB
* Final size of the binary project file
TOT size
SPOT/IKONOS
462 / 912 MB
177 / 268 MB *
459 / 904 MB
4. TOURIST INFORMATION SYSTEM
A GIS is a software used to map, collect, manage and analyze
geographical and other related kinds of attribute data. GIS
records the geometry and location of features in layers and the
combination of layers of information related to a place gives a
better understanding of the area. The GIS technology allows
database operations like queries and statistical analysis together
with the visualization and geographic analysis of maps.
The use of GIS in heritage management has been also
underlined by UNESCO, as a GIS allows (1) historical and
physical site documentation, (2) the assessment of physical
condition, cultural significance and administrative context, (3)
the preparation of conservation and management strategies, (4)
the implementation, monitoring and evaluation of management
policies [http://www.unescobkk.org/culture/gis/index.html].
Furthermore, a GIS tool generates permanent records of
heritage sites, including also text documentation, virtual flightover and 3D models.
The information recovered from the high-resolution satellite
imagery is imported in GIS software (ArcView and ArcGIS) for
further analysis, data visualization and topographic information
generation.
The package works with its own internal file format (IMG),
reducing considerable the file size. The LOD can be only
statically applied to the whole scene. A dynamic LOD can be
applied generating a ‘Virtual World’ (as it is called in the
software): the model is divided in different sectors and, for each
sector, the software constructs a pyramid of detail for the
geometry and the texture data. In the generated virtual world,
the user can easily navigate through the scene while the details
are automatically adapted according to the model’s distance
from the viewer.
3.2 Skyline–Terra Builder
The software converts all the data (geometry and texture) in its
own internal and binary file format (MPT), which can reduce
the total data size by a factor 4. The interactive and offline
navigation is very easy, due to the dynamic LOD.
Unfortunately different aliasing artefacts are present in the
rendered model and the LOD resampling produces a noticeable
reduction in the image texture quality, even in the foreground.
3.3 ESRI ArcGIS – ArcScene 8.3
ArcScene is mainly a 3D GIS tool of ArcGIS. The DTM data is
imported as regular grid and then a TIN is generated in an
internal binary format. The software cannot perform real-time
rendering of big data sets, it has a static LOD control but it can
easily display different vector information without aliasing
Figure 12: A view of the GIS data edited in ArcView.
The ‘cultural landscape and archaeological remains of the
Bamiyan
valley’,
as
defined
by
UNESCO
[http://www.whc.unesco.org/sites/208rev.htm]
includes
8
locations, identified with an area of interest and a buffer area:
1. Bamiyan Cliff including niches of the 38 meter Buddha,
seated Buddhas, 55 meter Buddha and surrounding caves;
2. Kakrak Valley caves including the niche of the standing
Buddha;
3. Qoul-I Akram Caves in the Fuladi Valley;
4. Kalai Ghamai Caves in the Fuladi Valley;
5. Shahr-i-Zuhak;
6. Qallay Kaphari A;
7. Qallay Kaphari B;
8. Shahr-i-Ghulghulah.
Man-made objects (e.g. streets and buildings) and rivers were
measured manually from the ortho-images. A total of 243
objects were extracted and then overlapped onto the recovered
DTM and ortho-image, using the extrude function of ArcScene
(Figure 13). For visualization reasons, an exaggeration of 5 m
was given to the extracted buildings.
From the DTM raster grid, the contour lines of the area can be
computed and displayed together with the man-made objects
for the new cartography of the Bamiyan village. Until now only
Russian maps from the 70’s are available, but, as shown in
Figure 14, the morphology of the village is changed.
In addition, the DTM data has been imported into a GIS
software to create a complete information system of the
UNESCO protected area, which may serve for technical,
scientific and touristic purposes in the future.
Figure 14: The new cartography of the Bamiyan village
generated from satellite imagery (below) compared with an old
aerial Russian map (above). Two big changes are signalised:
enlargement of the airport and the shift of the city-bazaar.
ACKNOWLEDGEMENTS
Figure 13: Two views (rendered with ArcScene) of the 3D
model of the Bamiyan area with the extracted man-made
structures (streets, houses and airport) and the rivers.
With other GIS functions and tools, the 8 protected locations
inside the Bamiyan area can be defined (according to their
buffer area), and displayed (Figure 15).
Now that these maps are available, further analysis can be
performed for administrative, conservation and monitoring
management.
5. CONCLUSIONS
In this work we have shown the modeling of the whole cultural
heritage site of Bamiyan, Afghanistan, from high-resolution
satellite imagery. We have used different types of sensors and
produced a detailed terrain model that will be used for
visualization, animation and documentation. The satellite
images were the only possibility for the mapping of the heritage
area because of their instant availability and high resolution.
For the photo-realistic 3D digital models of the entire Bamiyan
area different commercial packages have been used, but the
management of large terrain data sets is still problematic, in
particular for real-time rendering.
The authors would like to thank Daniela Poli for her help in the
satellite image acquisition, CNES for providing the SPOT5/HRS images at special conditions through the ISIS program
and Space Imaging for providing a IKONOS scene for free. We
also appreciate the contributions of Natalia Vassilieva in terms
of doing photogrammetric measurements.
REFERENCES
Gruen, A., 1985: Adaptive Least Squares Correlation: A
powerful Image Matching Technique. South Africa Journal of
P RS and Cartography, 14 (3), pp. 175-187.
Gruen, A. Remondino, F. Zhang, L., 2004a: 3D modeling and
visualization of large cultural heritage sites at very high
resolution: the Bamiyan valley and its standing Buddhas.
IAPRS&SIS, Vol. XXXV-B5.
Gruen, A. Remondino, F. Zhang, L., 2004b: Photogrammetric
Reconstruction of the Great Buddha of Bamiyan, Afghanistan
The Photogrammetric Record, Vol.19, No.107, pp. 177-199.
Poli, D., Zhang, L., Gruen, A., 2004: SPOT-5/HRS Stereo
Images Orientation and Automated DSM Generation.
IAPRS&SIS, Vol.XXXV-B1.
Zhang, L., Gruen, A., 2004: DSM Generation from Linear
Array Imagery. IAPRS&SIS, Vol.XXXV-B3.
Figure 15: The UNESCO protected areas near the Bamiyan village, with their buffer (blue) zone, displayed within ArcScene.
Kalai Ghamai Caves
Fuladi Valley
Qoul-I Akram
Shahr-i-Ghulghulah
The Bamiyan cliff with the
two empty niches and surrounding caves
The Kakrak empty niche
The Kakrak valley